OPTIMIZATION OF UHF SPECTRUM ALLOCATION LEVERAGING DEEP LEARNING

The study addresses spectrum scarcity and congestion caused by static spectrum management and the deployment of Internet of Things (IoT). To address these issues, this study proposes a Deep Learning based framework for Dynamic Spectrum Access (DSA) using Deep Neural Network (DNN) optimized with the Levenberg-Marquardt algorithm to solve an adaptive multi-objective spectrum allocation problem. The DNN treats allocation as a classification task, effectively mapping interference patterns and user demands as real-time channel assignments, unlike conventional heuristic approaches that depend on recurrent search. The model focused on five secondary users and seven frequency bands, to maximize throughput, spectral efficiency, and fairness while minimizing interference. The simulation results demonstrated a Spectral Efficiency of 10 bps/Hz, a Throughput of 1.5*107 bps, a nearly flawless Fairness Index of 1.00 and attaining minimal interference of 90 dB, outperforming current state-of-the-art techniques and showcasing its potential for reliable and fair spectrum management in next-generation cognitive radio networks.